Skip to content

Commit

Permalink
feat: updating the smax plots
Browse files Browse the repository at this point in the history
  • Loading branch information
OmaymaMahjoub committed Feb 8, 2024
1 parent decd959 commit a9dc878
Show file tree
Hide file tree
Showing 8 changed files with 1 addition and 1 deletion.
Binary file modified docs/images/smax_results/10m_vs_11m.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/27m_vs_30m.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/3s5z.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/3s5z_vs_3s6z.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/3s_vs_5z.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/5m_vs_6m.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
Binary file modified docs/images/smax_results/6h_vs_8z.png
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
2 changes: 1 addition & 1 deletion docs/smax_benchmark.md
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
# StarCraft Multi-Agent Challenge in JAX

We trained Mava’s recurrent systems on eight varied SMAX scenarios, We executed a tuning sweep to identify the optimal hyperparameters for each scenario, followed by a benchmarking. The outcomes were then compared to the final win rates reported in [Rutherfordet al., 2023](https://arxiv.org/pdf/2311.10090.pdf), To ensure equitable comparisons, we adhered to specific criteria, including training over 10 million timesteps, utilising 64 parallel environments, and employing 10 different seeds.
We trained Mava’s recurrent systems on eight varied SMAX scenarios, We executed a tuning sweep to identify the optimal hyperparameters for each scenario, followed by a benchmarking. The outcomes were then compared to the final win rates reported in [Rutherfordet al., 2023](https://arxiv.org/pdf/2311.10090.pdf). To ensure equitable comparisons, we adhered to specific criteria, including training over 10 million timesteps, utilising 64 parallel environments, and employing 10 different seeds.

Please see below for Mava's recurrent IPPO and MAPPO performance on these scenarios.
</br>
Expand Down

0 comments on commit a9dc878

Please sign in to comment.